JPWO2019181969A1 - Imaging device and imaging method - Google Patents

Abstract

Provided is an imaging device or the like capable of improving the speed of imaging processing. The image pickup device has a region through which the two-dimensional signal indicating the image is transmitted, and a filter provided with a portion that blocks at least a part of the two-dimensional signal in the region and the power of the two-dimensional signal transmitted through the filter. The image represented by the two-dimensional signal is reproduced based on the plurality of powers detected when the positional relationship between the detector to be detected and the distribution of the portion that blocks at least a part of the two-dimensional signal is different from each other. It is equipped with a reconstruction means to be constructed.

Description

The present disclosure relates to an imaging apparatus and an imaging method.

A technique called compressed sensing, which restores an object such as an image from observation data that is less than the required data, is being developed. In compressed sensing, as an algorithm for restoring an object, for example, an algorithm such as the ADMM (alternation direction method of multipliers) algorithm described in Non-Patent Document 1 is used.

S. Boyd, N.M. Parikh, E.I. Chu et al, Distributed Optimization and Static Learning via the Alternation Direction Machine Learning of Multipliers, Foundation and Trends3, Funding and Trends in

Imaging of images and the like is generally preferably performed at high speed. Therefore, even when the compressed sensing technology is used, it is desired to speed up the imaging process including the acquisition of observation data and the construction of an image from the acquired observation data.

A main object of the present disclosure is to provide an imaging device or the like capable of improving the speed of imaging processing in order to solve the above problems.

The image pickup apparatus according to one aspect of the present disclosure includes a filter having a region through which a two-dimensional signal indicating an image is transmitted, and a portion for blocking at least a part of the two-dimensional signal in the region through which the two-dimensional signal is transmitted. , Detected when the positional relationship between the detector that detects the power of the two-dimensional signal that has passed through the filter and the distribution of the two-dimensional signal that is imaged on the filter and the part that blocks at least a part of the two-dimensional signal is different. It is provided with a reconstruction means for reconstructing the image represented by the two-dimensional signal based on the plurality of powers generated.

Further, the imaging method according to one aspect of the present disclosure has a region through which the two-dimensional signal indicating the image is transmitted, and a portion for blocking at least a part of the two-dimensional signal is provided in the region through which the two-dimensional signal is transmitted. Multiple powers detected when the power of the two-dimensional signal transmitted through the filter is detected and the positional relationship between the two-dimensional signal imaged on the filter and the distribution of the portion that blocks at least a part of the two-dimensional signal is different. Based on, the image represented by the two-dimensional signal is reconstructed.

According to the present disclosure, it is possible to provide an imaging device or the like capable of improving the speed of imaging processing.

It is a figure which shows the image pickup apparatus in an embodiment. It is a figure which shows the detailed configuration example of the image pickup apparatus in embodiment. It is a figure which shows the structural example at the time of obtaining the transformation matrix used in the image pickup apparatus in embodiment. It is a figure which shows the relationship between a filter pattern and a transformation matrix. It is a figure which shows the example of the image to be reconstructed. It is a figure which shows the example of the pattern of the filter used at the time of the observation of power. It is a figure which shows the example which changed the positional relationship between the two-dimensional signal shown in FIG. 5A, and the pattern of a filter. It is a figure which shows the configuration example when the change of the positional relationship between a two-dimensional signal and a pattern of a filter is controlled by a pattern control unit. It is a figure which shows an example of the case where the positional relationship between a two-dimensional signal and a filter pattern changes. It is a figure which shows an example of the case where the positional relationship between a two-dimensional signal and a filter pattern changes. It is a figure which shows an example of the case where the positional relationship between a two-dimensional signal and a filter pattern changes. It is a flowchart which shows the operation of the image pickup apparatus in embodiment. It is a figure which shows an example of the image pickup apparatus in the modification of embodiment. It is a figure which shows the example of the original image in the simulation example. It is a figure which shows the example of the image reconstructed in the simulation example.

The embodiments of the present disclosure will be described with reference to the drawings. FIG. 1 is a diagram showing an imaging device according to an embodiment.

As shown in FIG. 1, the image pickup apparatus 100 in the embodiment includes a filter 110, a detector 120, and a reconstruction unit 130. The filter 110 has a region through which a two-dimensional signal showing an image is transmitted, and a portion for blocking at least a part of the two-dimensional signal is provided in the region through which the two-dimensional signal is transmitted. The detector 120 detects the power of the two-dimensional signal that has passed through the filter 110. The reconstruction unit 130 is based on the power of a plurality of two-dimensional signals detected when the positional relationship between the two-dimensional signal and the distribution of the portion blocking at least a part of the two-dimensional signal in the filter 110 is different. Reconstruct the image represented by the dimensional signal.

In the following embodiments, the two-dimensional signal is, for example, an image of light such as visible light imaged by a lens. That is, the two-dimensional signal is a signal that is converged by an element such as a lens that converges the signal, and represents some image on the surface to be imaged as a whole or a part thereof.

As a more specific example, the image pickup apparatus 100 includes the configuration shown in FIG. In the configuration shown in FIG. 2, in addition to the above-mentioned elements, the image pickup apparatus 100 further includes a pattern control unit 111, a lens 11, an iris 12, and a scattering medium 13. In the configuration shown in FIG. 2, it is assumed that the two-dimensional signal is, for example, visible light.

The image reconstructed by the image pickup apparatus 100 is obtained by the lens 11. The image obtained by the lens is imaged on the filter 110. That is, the filter 110 is usually arranged at a position where the image obtained by the lens 11 is imaged. Further, the iris 12 and the scattering medium 13 are provided as needed in order to reconstruct the image with higher accuracy. The iris 12 blocks miscellaneous signals. The scattering medium 13 homogenizes the signal that has passed through the filter 110.

Further, the pattern control unit 111 controls the filter 110 so as to change the positional relationship between the two-dimensional signal and the distribution of the portion of the filter 110 that blocks at least a part of the two-dimensional signal. Hereinafter, in the present embodiment, the distribution of the portion of the filter 110 that blocks at least a part of the two-dimensional signal may be referred to as a pattern of the filter 110 or simply a pattern.

The image pickup apparatus 100 reconstructs an image by using the compressed sensing technique described above. More specifically, it passes through the pattern of the filter 110 and reconstructs the image represented by the signal based on the power of the signal detected by the detector 120. The number of powers detected by the detector 120 is less than the number of pixels in the reconstructed image.

The image reconstruction is performed according to the ADMM algorithm, which is one of the compressed sensing techniques, using the transformation matrix D obtained according to the pattern of the filter 110. Hereinafter, a procedure for obtaining the transformation matrix D and a procedure for reconstructing an image based on the ADMM algorithm using the transformation matrix D will be described.

In the following description of the procedure for determining the transformation matrix and the procedure for reconstructing the image, it is assumed that the two-dimensional signal is visible light. Further, it is assumed that the reconstructed image is a general visible light image. However, as will be described later, the two-dimensional signal is not limited to visible light, and may be a signal such as an electromagnetic wave other than visible light or a sound wave including ultrasonic waves.

First, the procedure for obtaining the transformation matrix D will be described. The transformation matrix D is obtained by using the configuration shown in FIG. 3 as an example. The number of rows and columns of the transformation matrix D is determined according to the number of pixels of the image to be reconstructed. In the following, it is assumed that the image to be reconstructed is x, and the image x is an image having m pixels in the horizontal direction and n pixels in the vertical direction. Hereinafter, the procedure for obtaining the transformation matrix D may be referred to as a calibration step or simply calibration.

In the example shown in FIG. 3, the signal generated from the signal generator 14 passes through the filter 110. Then, the light transmitted through the filter 110 is detected by the two-dimensional signal detector 15. The filter 110 has a region on which a two-dimensional signal is transmitted on a surface facing the signal generator 14 or the two-dimensional signal detector 15. Then, in this region of the filter 110, a portion for blocking at least a part of the two-dimensional signal is provided. As a portion of the filter 110 that blocks at least a part of the two-dimensional signal, a semitransparent portion in which a part of the two-dimensional signal is blocked and a part is transmitted, and a non-transparent portion in which the two-dimensional signal is blocked and not transmitted are included. included. Further, it is assumed that the part other than the part that blocks at least a part of the two-dimensional signal is the part through which the two-dimensional signal is transmitted as it is. Details of these will be described later.

Further, in the example shown in FIG. 3, the region through which the two-dimensional signal of the filter 110 is transmitted is preferably arranged so as to face a direction perpendicular to the traveling direction of the signal generated from the signal generator 14.

The signal generator 14 generates a two-dimensional signal corresponding to the signal detected by the detector 120. The signal generator 14 preferably generates a signal having a wavelength close to the wavelength of the two-dimensional signal detected by the detector 120. For example, when the two-dimensional signal is visible light, the signal generator 14 generates visible light.

The signal generator 14 is not particularly limited in its specific configuration as long as it can generate a desired signal according to the wavelength of visible light, electromagnetic waves, or the like detected by the detector 120. However, it is preferable that the signal generator 14 can uniformly irradiate the pattern of the filter 110, that is, the difference in power of the signals radiated to each region of the pattern is as small as possible.

The two-dimensional signal detector 15 measures the power of the signal generated from the signal generator 14 and passed through the filter 110. The two-dimensional signal detector 15 detects the power of each signal position, unlike the detector 120 whose details will be described later. Therefore, as the two-dimensional signal detector 15, a sensor having a pixel count equal to or greater than the pixel count of the image reconstructed by the image pickup apparatus 100 is used. The sensitivity characteristic of the two-dimensional signal detector 15 is preferably close to the sensitivity characteristic of the detector 120.

When the signal is visible light, a camera of a general CCD (Charge Coupled Device) or CMOS (Complementary metal-axis-ometry) image sensor is used as the two-dimensional signal detector 15. Further, as the two-dimensional signal detector 15, for example, an ultraviolet camera if the signal is ultraviolet rays, an InGaAs (indium gallium arsenic) camera if the signal is near infrared rays, a thermal image camera if the signal is mid-far infrared rays, or the like. Used. The arrangement of pixels including the number of pixels and the aspect ratio of the two-dimensional signal detector 15 is determined according to the number of pixels and the arrangement of pixels of the image to be reconstructed.

The power of the signal generated from the signal generator 14 and transmitted through the filter 110 changes for each pixel according to the positional relationship between the pixels of the two-dimensional signal detector 15 and the above-mentioned pattern of the filter 110. That is, the two-dimensional signal detector 15 detects the power of the signal from the signal generator 14, which has changed for each pixel according to the positional relationship between the pixels of the two-dimensional signal detector 15 and the filter 110 pattern.

Further, the power of the signal detected by the two-dimensional signal detector 15 changes the positional relationship between the pattern of the filter 110 and the pixels of the two-dimensional signal detector 15 due to the rotation of the filter 110 or the like. It changes depending on the situation. The transformation matrix D is obtained based on the plurality of powers obtained as described above.

The procedure for obtaining the transformation matrix D using the configuration shown in FIG. 3 will be further described with reference to FIG. FIG. 4A schematically shows an example of the pattern of the filter 110. In the example shown in FIG. 4A, in the pattern of the filter 110, the region where the signal is blocked and does not pass through is represented by a black square, and the region where the signal passes through as it is is represented by a white square in units of pixels. ..

In this case, first, the signal generator 14 generates a signal, and the two-dimensional signal detector 15 detects the signal that has passed through the filter 110. The power of the signal detected by the two-dimensional signal detector 15 has a different magnitude for each pixel depending on the distribution of the portion of the filter 110 that blocks at least a part of the two-dimensional signal.

In the example shown in FIG. 4A, the signal power for each pixel detected by the two-dimensional signal detector 15 is a relatively large value for the pixels in the region where the two-dimensional signal is transmitted in the filter 110. .. Similarly, the signal power for each pixel detected by the two-dimensional signal detector 15 is a relatively small value for the pixel at the portion where at least a part of the two-dimensional signal is blocked by the filter 110. Then, the power of the signal for each pixel detected by the two-dimensional signal detector 15 is arranged as shown by the arrow shown in FIG. 4 (A), and is m × as shown in D ₁ of FIG. 4 (B). It is represented as an n-dimensional vector.

In the example shown in FIG. 4B, the power of the signal is represented as a binary vector of m × n dimensions. The element of the vector corresponding to the pixel in which the power of a large value is detected is represented by a white square, and the element of the vector corresponding to the pixel in which the power of a small value is detected is represented by a black square.

Such power detection for each pixel is repeated for each case where the positional relationship between the pixels of the two-dimensional signal detector 15 and the pattern of the filter 110 is different. The positional relationship between the pixels of the two-dimensional signal detector 15 and the pattern of the filter 110 is changed, for example, by changing the positional relationship between the pixels of the two-dimensional signal detector 15 and a specific pattern of the filter 110. May be good. Further, the change in the positional relationship between the pixels of the two-dimensional signal detector 15 and the pattern of the filter 110 may be performed by using a filter 110 having a different distribution of portions that block at least a part of the two-dimensional signal.

The change in the positional relationship between the pixels of the two-dimensional signal detector 15 and the specific pattern of the filter 110 is performed, for example, by rotating the filter 110 about the direction in which the signal is transmitted. Further, the change in the positional relationship between the pixels of the two-dimensional signal detector 15 and the specific pattern of the filter 110 may be performed by moving the filter 110 in a direction intersecting the direction in which the signal is transmitted.

When the positional relationship between the pixels of the two-dimensional signal detector 15 and the pattern of the filter 110 changes, the positional relationship between the signal generator 14, the filter 110, and the two-dimensional signal detector 15 (the signal in FIG. 3 advances). Do not change the distance in the direction, etc.).

FIG. 4A schematically shows an example in which the positional relationship between the pixels of the two-dimensional signal detector 15 and the pattern of the filter 110 is changed. Then, in the power detection that is repeatedly performed by changing the positional relationship between the pixels of the two-dimensional signal detector 15 and the pattern of the filter 110, the power of the signal for each pixel detected by the two-dimensional signal detector 15 is used. Multiple vectors are required. That is, D _k from k vectors D ₁ obtained by performing the positional relationship by changing the k times of power observed between the pattern of pixels and the filter 110 of the two-dimensional signal detector 15, the transform matrix D Is configured. That is, each row of the transformation matrix D is a vector D _i (1 ≦ i ≦ k) obtained for a specific positional relationship with respect to the pixel with the two-dimensional signal detector 15 of a certain pattern of the filter 110.

Next, the procedure for reconstructing the image based on the ADMM algorithm will be described. In this case, the transformation matrix D obtained by the above procedure is used. Further, in the following procedure, it is assumed that the image x to be reconstructed is an image having m pixels in the horizontal direction and n pixels in the vertical direction as described above.

In the image pickup apparatus 100, a plurality of powers of two-dimensional signals transmitted through the pattern of the filter 110 and detected by the detector 120 are used. The power of the plurality of two-dimensional signals is the power detected by the detector 120 by changing the positional relationship between the two-dimensional signal and the distribution of the portion of the filter 110 that blocks at least a part of the two-dimensional signal. is there.

When the image is reconstructed by the image pickup apparatus 100, the power of the two-dimensional signal is observed k times by setting the filter 110 in the same state as in the case of obtaining the transformation matrix D. That is, when the detector 120 detects the power of the two-dimensional signal, the filter 110 having the same pattern as that for obtaining the transformation matrix D is used. Further, in each of the k power observations of the two-dimensional signal, the positional relationship between the filter 110 and the two-dimensional signal is the positional relationship between the two-dimensional signal of the filter 110 and the two-dimensional signal detector 15 and the corresponding pixel. It has the same positional relationship.

An example in which the power is observed k times will be described with reference to FIGS. 5A, 5B, and 5C. FIG. 5A shows an example of a reconstructed image. The light rays representing this image are converged by the lens 11. FIG. 5B shows an example of the pattern of the filter 110 used when observing the power. FIG. 5C shows an example in which the positional relationship between the two-dimensional signal shown in FIG. 5A and the pattern of the filter 110 is changed. That is, when the image representing the image shown in FIG. 5A is reconstructed using the filter 110 shown in FIG. 5B, as an example, as shown in FIG. 5C, the positional relationship between the two-dimensional signal and the pattern of the filter 110 is determined. The power is observed by changing it.

In this embodiment, the detector 120 is a detector that does not detect position information such as the power distribution of a two-dimensional signal, but only detects the power of the signal. Therefore, the power of the two-dimensional signal detected by the detector 120 is the sum of the signal intensities of all the pixels of the image to be reconstructed when passing through the filter 110.

In this case, the image x is reconstructed by the imaging device 100, a power j _i of the signal detected by detector 120 with respect to the pattern corresponding to the vector D _i constituting one of the rows of the transform matrix D , It is expressed by the relationship as shown in the following equation (1).

(1) In the formula, d ⁱ _k denotes the (m × n) of elements in the vector D _i as described above, x _k is a value representing the intensity of each of the m × n pixels included in the image x Represents the values arranged in order. d ⁱ _k and _{x k} are values at corresponding positions in the respective images.

It is assumed that the detection of the signal power by the detector 120 is performed on the patterns corresponding to each of D ₁ to D _k corresponding to any row of the transformation matrix D. In this case, the relationship of the following equation (2) can be obtained with respect to the vector J for the power j _i of the k signals detected for each of D ₁ to D _k .

For each of D, x and J in Eq. (2), the subscript indicates the number of elements in the matrix representing each. D _{k × mn} is a transformation matrix of k rows × (m × n) columns. That is, the transformation matrix D _{k × mn} is composed of k vectors D _i (i = 1 to k) including (m × n) elements, respectively. x _{mn × 1} is a vector of (m × n) rows. J _{k × 1} is a vector of k rows. In the following description, each of D, x and J without a subscript represents the same content as above.

In compressed sensing, a solution to the minimization problem shown in Eq. (3) below is required to reconstruct an image. That is, by obtaining the L1 norm solution of Eq. (3), the image x is reconstructed from the detection result of the power of the k signals described above.

In equation (3), || x || ₁ represents the L1 norm of x.

The above solution is obtained using the ADMM algorithm. First, consider the cost function expressed by the following equation (4).

Equation (4) is a cost function in the Lagrange undetermined multiplier method. Further, in equation (4), ν represents a Lagrange undetermined multiplier.

In equation (3), a new variable z is introduced to distinguish x for the L1 norm from other x. The problem of minimizing L (x) shown in Eq. (4) is replaced by the conditional minimization problem shown in Eq. (5) below. In equation (5), λ represents the cost coefficient of the Lagrange undetermined multiplier method.

Then, according to the procedure of the extended Lagrange method, the new cost function shown in the following equation (6) is minimized.

In equation (6), u [t] indicates an auxiliary term that converges to the optimum solution when the constrained optimization problem is solved by the gradient method that iteratively calculates from an appropriate initial point. When the equation (6) is differentiated with respect to x, the following equation (7) is obtained.

The following equation (8) is obtained, where x is set to equation (7).

Substituting x obtained as in Eq. (8) into the original Eq. (6), Eq. (6) becomes as in Eq. (9) below.

Equation (9) can be considered to be a quadratic function with respect to ν. Therefore, the optimum solution that minimizes the equation (9) is expressed by the following equation (10).

By substituting the obtained equation (10) into the equation (8), the following equation (11) is obtained.

ただし

It is obtained by the gradient method for iterative calculation described above. When the gradient method is applied to the equation (11), the following equation (12) is obtained as an equation showing the t + 1th value of the iterative calculation for x, z and u.

However

In this embodiment, it is assumed that the iterative calculation is performed only once. In this case, z [t + 1] and u [t + 1] do not need to be considered. Further, although it is necessary to initialize z [t] and u [t], z [0] and u [0] may be 0. Therefore, the following equation (13) is obtained as the image x to be reconstructed.

On the other hand, the ADMM algorithm is very effective when a basis that can be expected to be sparsity is found through a certain transformation. When a signal is sparse, it means that many components of the signal are zero. Therefore, in reconstructing the image x, it is generally necessary to convert x into a space having sparsity. That is, the image is reconstructed in a space having sparsity.

Sparsification is realized, for example, by performing a discrete Fourier transform or a wavelet transform on the image x. Therefore, as shown in the following equation (14), the image x is sparsed by using the sparse transformation matrix Φ. The sparse transformation matrix Φ is, for example, either a discrete Fourier transform matrix or a wavelet transformation matrix. Then, Q is obtained by the ADMM algorithm described above.

Equation (14) is transformed into the form shown in equation (15) below using the inverse matrix Φ ^-1 of the sparse transformation matrix Φ.

Further, the equation (15) is further converted into the form shown in the following equation (16) by using the relation shown in the equation (2).

When sparse transformation matrix [Phi is the discrete Fourier transform or wavelet transform matrix represents the conjugate transpose matrix of [Phi subscript +, if the [Phi ^+, the [Phi and [Phi ⁺ The inverse to each other.
Therefore, the following equation (17) can be obtained from the equation (16).

In equation (17), the unknown is Q. Therefore, the approximate solution Q ”of Q is obtained by obtaining the minimum solution of the L1 norm as shown in the following equation (18).

With respect to the equation (18), the following equation (19) can be obtained in the same manner as in the example in which the equation (16) is obtained with respect to the equation (3) described above.

In equation (19), P represents (D' _{k × mn}・ Ф ⁺ ) ⁺・ inv [(D' _{k × mn}・ Ф ⁺ ) ・ (D' _{k × mn}・ Ф ⁺ ) ⁺ ].

That is, the image x to be reconstructed can be obtained by obtaining Q "which is an approximate solution of Q and using the obtained Q" and the above-mentioned Ф ⁺ .

The number of observations k by the two-dimensional signal detector 15 or the detector 120 may generally be smaller than m × n, which is the number of pixels of the reconstructed image x. For example, when the above-mentioned sparsification is appropriately performed, the number of observations by the two-dimensional signal detector 15 or the detector 120 may be about a few percent (percentage) of the number of pixels. That is, the image pickup apparatus 100 can reconstruct an image from a small amount of observation data obtained by changing the positional relationship between the two-dimensional signal and the pattern of the filter 110 by using the ADMM algorithm which is a method of compressed sensing. And.

The number of observations k by the two-dimensional signal detector 15 or the detector 120 is not limited to the above-mentioned example, and is appropriately determined according to the image quality required for the reconstructed image and the degree of sparsification.

Subsequently, the details of each element of the image pickup apparatus 100 in the present embodiment will be described.

The lens 11 forms an image reconstructed by the image pickup apparatus 100. The lens 11 is arranged on the filter 110 so as to form an image.

When the two-dimensional signal is visible light, the lens 11 is, for example, a general optical lens. Although the lens 11 is shown as a single lens in FIG. 2, the lens 11 is not limited to this. The lens 11 may be composed of a plurality of lenses. The specific configuration of the lens 11 is appropriately determined according to conditions such as the configuration of the filter 110 and the configuration or type of the detector 120.

Further, when the two-dimensional signal is an electromagnetic wave or a sound wave other than visible light, an appropriate configuration for converging the two-dimensional signal is used as the lens 11 according to the type of signal. For example, a reflector may be used as the lens 11.

As described above, the iris 12 and the scattering medium 13 are provided as necessary for the reconstruction of the image with higher accuracy by the image pickup apparatus 100.

The iris 12 blocks miscellaneous signals included in the signal collected by the lens. As an example, when the two-dimensional signal is visible light, a general optical lens diaphragm mechanism is used as the iris 12.

Further, the scattering medium 13 disperses the off-focus signal that is not converged by the lens 11 among the signals directed to the detector 120. By providing the scattering medium 13, the two-dimensional signal converged by the lens 11 is made uniform with respect to the direction toward the detector 120. As the scattering medium 13, for example, frosted glass, a lump of fine particles such as alumina, or the like is used depending on the type of the two-dimensional signal.

Further, as shown in FIG. 2, the scattering medium 13 is preferably brought into close contact with the surface on which the detector 120 detects a signal. With such a configuration, the loss of power due to the reflection of the two-dimensional signal from the detection surface of the detector 120 is reduced.

By providing the iris 12 and the scattering medium 13, it is possible to remove noise contained in the signal detected by the detector 120, reduce the power loss of the two-dimensional signal detected by the detector 120, and the like. As a result, the detection accuracy of the power of the two-dimensional signal detected by the detector 120 is improved. Then, the image can be reconstructed with higher image quality.

Each of the iris 12 and the scattering medium 13 is provided as needed. One or both of the iris 12 and the scattering medium 13 may be provided depending on the type of the two-dimensional signal to be detected, the relationship with other components, the image quality required for the reconstructed image, and the like. It does not have to be.

The filter 110 has a region through which a two-dimensional signal is transmitted. Then, in the region of the filter 110, a portion that blocks at least a part of the two-dimensional signal (that is, a portion that has a different transmittance of the two-dimensional signal from other portions) is provided. In the image pickup apparatus 100, the filter 110 is arranged at a position where an image is formed by the lens 11.

In the present embodiment, the distribution of the portion of the filter 110 that blocks at least a part of the two-dimensional signal is not particularly limited. That is, in the filter 110, a portion having an arbitrary transmittance is provided at an arbitrary place. The portion of the filter 110 that blocks at least a part of the two-dimensional signal may be provided corresponding to the pixel of the image to be reconstructed, or may be provided independently of the pixel.

The portion of the filter 110 that blocks at least a part of the two-dimensional signal and its distribution will be further described.

In the region where the two-dimensional signal of the filter 110 is transmitted, the two-dimensional signal is transmitted without being blocked at a portion other than a portion that blocks at least a part of the two-dimensional signal. That is, in a portion other than a portion that blocks at least a part of the two-dimensional signal, the two-dimensional signal passes through the filter 110 at a transmittance according to the material of the filter 110.

The portion of the filter 110 that blocks at least a part of the two-dimensional signal includes a non-transmissive portion or a semi-transparent portion. In the non-transmissive portion, the two-dimensional signal may be blocked and not transmitted, or a portion having a transmittance low enough to be treated as substantially non-transparent of the two-dimensional signal may be regarded as the non-transmissive portion. At the semitransparent site, part of the two-dimensional signal is blocked and the rest is transmitted. The transmittance of the two-dimensional signal is not particularly limited in the semitransparent portion. Further, when a plurality of semitransparent parts are included as the parts that block at least a part of the two-dimensional signal in the filter 110, the transmittance of the two-dimensional signal at each part may be the same. , May be different. By providing a semi-transmissive portion so that the transmittance has more stages within the range corresponding to the sensitivity resolution of the detector 120, it is possible to reconstruct an image with less noise.

Further, the distribution of the portion of the filter 110 that blocks at least a part of the two-dimensional signal may have periodicity, a pseudo-random distribution, or a random distribution without periodicity.

If the distribution of the portion of the filter 110 that blocks at least a part of the two-dimensional signal has periodicity, noise corresponding to the period may appear in the reconstructed image. Therefore, the change in transmittance in the pattern is most preferably a completely random change, and at least a pseudo-random change expressed in a form other than the superposition of simple analysis formulas. The superposition of simple analysis formulas includes, for example, the sum or product of trigonometric functions having several kinds of frequencies, or the linear addition of these. Since the change in transmittance in the pattern is random or pseudo-random, it is possible to reconstruct an image with less noise as compared with the case where the change in transmittance has periodicity.

When the two-dimensional signal is visible light, the filter 110 is made of a material such as glass whose transmission amount can be adjusted with respect to light. A pattern is provided by forming a pattern on the surface of the glass that reduces the transmittance by an appropriately selected method. When the two-dimensional signal is an electromagnetic wave other than visible light, a metal material or the like is used as the filter 110. When the two-dimensional signal is a sound wave, a material that changes the acoustic impedance, such as a metal mask, is used as the filter 110.

Further, in the image pickup apparatus 100, as described above, the image is reconstructed based on the power detected by changing the positional relationship between the two-dimensional signal and the pattern of the filter 110. The change in the positional relationship between the two-dimensional signal and the pattern of the filter 110 is performed by changing the positional relationship between the two-dimensional signal, the specific pattern of the filter 110, and the two-dimensional signal. Therefore, the pattern control unit 111 controls the position of the filter 110 so that the positional relationship between the two-dimensional signal and the pattern of the filter 110 changes.

As described above, the change in the positional relationship between the two-dimensional signal and the pattern of the filter 110 is performed, for example, by rotating the filter 110 about the direction in which the two-dimensional signal is transmitted. Further, the change in the positional relationship between the two-dimensional signal and the pattern of the filter 110 may cause the filter 110 to be moved in any in-plane direction on the plane of the filter through which the signal is transmitted. For example, the direction in which the signal is transmitted is the optical axis direction, and when the optical axis direction is the z direction, the change in the positional relationship with the pattern is an arbitrary direction in the xy plane. Therefore, the pattern control unit 111 controls to change the position of the filter 110 so as to change the positional relationship between the two-dimensional signal and the pattern of the filter 110.

The positional relationship between the two-dimensional signal and the pattern of the filter 110 by the pattern control unit 111 is the same as the positional relationship between the pattern of the filter 110 and the pixels of the two-dimensional signal detector 15 when obtaining the transformation matrix D. Is controlled by.

FIG. 6 shows a configuration example in which the position of the filter 110 is controlled by the pattern control unit 111 so that the positional relationship between the two-dimensional signal and the pattern of the filter 110 changes. In the example shown in FIG. 6, the scattering medium 13 and the detector 120 are arranged so as to face one surface through which the two-dimensional signal of the filter 110 is transmitted. That is, in the example shown in FIG. 6, it is assumed that a two-dimensional signal is incident on the filter 110 from the side opposite to the surface of the filter 110 facing the scattering medium 13, and an image is formed on the surface on which the pattern of the filter 110 is drawn. To.

Further, in the example shown in FIG. 6, the filter 110 is provided with a rotating shaft 112. That is, the filter 110 is configured to be rotatable about the rotation shaft 112. Then, the pattern control unit 111 controls so that the positional relationship between the two-dimensional signal and the pattern of the filter 110 changes by rotating the filter 110 at the position where the two-dimensional signal is imaged by the lens 11.

The movement of the filter 110 is not limited to the example shown in FIG. An example in which the positional relationship between the two-dimensional signal and the pattern of the filter 110 changes due to the movement of the filter 110 will be described with reference to FIGS. 7A, 7B, and 7C. In each of FIGS. 7A, 7B, and 7C, a black-and-white pattern is a pattern provided on the filter 110. In this case, it is assumed that the black region of the filter 110 is the non-transparent region described above, and the white region of the filter 110 is the region through which the two-dimensional signal is transmitted as it is. Further, in each of FIGS. 7A, 7B, and 7C, it is assumed that the face is a reconstructed image.

In FIG. 7A, the pattern of the filter 110 rotates about the central portion of the face, which is the image to be reconstructed. Further, in FIG. 7B, the filter 110 is arranged so that the face, which is the image to be reconstructed, is located at a position deviated from the center of the pattern. Then, the pattern of the filter 110 rotates about the central portion of the pattern as an axis. Further, in FIG. 7C, the filter 110 moves in a direction along a surface provided with a pattern while rotating.

The pattern control unit 111 is not limited to the example shown in FIG. 6, and may control the position of the filter 110 as shown in FIGS. 7A, 7B, and 7C. The control of the position of the filter 110 by the pattern control unit 111 is not limited to these examples. Then, as shown in these examples, the detector 120 detects the power of the two-dimensional signal while rotating or moving the filter 110, so that the power of the plurality of two-dimensional signals required for image reconstruction is high. Can be detected.

As a method for controlling a specific position by the pattern control unit 111 and a mechanism for changing the position of the filter 110, a known method or mechanism is appropriately used.

Further, when the change in the positional relationship between the two-dimensional signal and the pattern has periodicity, noise corresponding to the periodicity may appear in the reconstructed image. Therefore, it is preferable that the change in the positional relationship between the two-dimensional signal and the pattern is a random change without periodicity or a change close to a random change, similar to the change in the transmittance in the pattern of the filter 110.

For example, in FIGS. 7A, 7B, and 7C, in the case shown in FIG. 7A, it is assumed that circumferential noise is generated around the central portion of the face, which is the center of rotation of the pattern. Further, in the example shown in FIG. 7B, it is assumed that fan-shaped noise is generated around the central portion of the rotation of the pattern. On the other hand, in FIG. 7C, noise is generated according to the moving direction of the pattern, but the generated noise is smaller than that of the example of FIG. 7A or FIG. 7B, or is less periodic and inconspicuous. Is assumed. Therefore, in the examples of FIGS. 7A, 7B, and 7C, in order to reduce or make the noise contained in the reconstructed image inconspicuous, the filter 110 is moved as shown in the example shown in FIG. 7C. Is preferable.

Further, as described above, the change in the positional relationship between the two-dimensional signal and the pattern of the filter 110 may be performed by using the filter 110 having a different distribution of the portions that block at least a part of the two-dimensional signal. For example, when the filter 110 is provided with a mechanism for changing a portion having a different two-dimensional signal transmittance, the mechanism blocks at least a part of the region through which the two-dimensional signal is transmitted of the filter 110. The distribution may be changed. Further, by using a plurality of filters 110, the positional relationship between the two-dimensional signal and the pattern of the filter 110 may be changed.

As described above, the detector 120 detects the power of the two-dimensional signal that has passed through the filter 110. As the detector 120, a general detector that detects the power of visible light or electromagnetic waves of a specific wavelength or wavelength band is appropriately used depending on the wavelength of the two-dimensional signal and other conditions.

In this embodiment, the detector 120 detects the magnitude of the power of the signal. That is, the detector 120 does not have to detect position information such as the distribution of signal power. The detector 120 may be a one-pixel sensor, and does not need to be a sensor that detects the position information of the power of the two-dimensional signal like the array-shaped sensor. The image pickup apparatus 100 makes it possible to acquire a two-dimensional image without using an array-shaped sensor, which may be expensive depending on the wavelength band to be detected.

As the detector 120, a general detector capable of detecting the power of the speckle pattern is appropriately used depending on the wavelength of the two-dimensional signal and other conditions. As the detector 120, a camera equipped with a CCD camera or a CMOS image sensor, an ultraviolet camera, an InGaAs camera, a thermal image camera, or the like is appropriately used depending on the wavelength of the two-dimensional signal and other conditions.

The gradation of the power of the signal that can be detected by the detector 120 is not particularly limited. When the sensitivity resolution of the detector 120 is high and the power of the two-dimensional signal is detected in more gradations, the filter 110 including the parts of the transmittance of many stages according to the sensitivity resolution can be used. , It is possible to reconstruct an image with less noise.

The reconstruction unit 130 reconstructs the image based on the powers of the plurality of signals detected by the detector 120 and using the transformation matrix D previously obtained as described above. Each of the plurality of signals detected by the detector 120 is the power detected when the distribution of the two-dimensional signal and the portion that blocks at least a part of the two-dimensional signal of the filter 110 are different.

More specifically, the reconstruction unit 130 obtains Q based on the power of the signal obtained by the detection by the detector 120 k times, using the relationship of the above equation (18). As described above, Q is a value obtained by subjecting the image x to a discrete Fourier transform or a wavelet transform. Then, when Q is obtained, the reconstruction unit 130 reconstructs the image using the inverse matrix Φ ^-1 of the sparse transformation matrix Φ. In the present embodiment, it is assumed that the image reconstructed in this case is a monochromatic image.

The reconstruction unit 130 is realized, for example, by appropriately combining hardware including a CPU (Central Processing Unit) and a memory, and software for reconstructing an image. The specific configuration of the reconstruction unit 130 is not particularly limited, and may be realized by an FPGA (Field Programmable Gate Array), dedicated hardware, or the like. The reconstruction unit 130 may include a function of obtaining the transformation matrix D by the procedure of the calibration step described with reference to FIG.

Subsequently, an example of the operation of the image pickup apparatus 100 will be described with reference to the flowchart shown in FIG. In the following description of the operation, it is assumed that the transformation matrix D is obtained in advance for the pattern of the filter 110 by the procedure of the calibration step described with reference to FIG.

First, a two-dimensional signal indicating an image transmitted through the lens 11 is imaged in a region through which the two-dimensional signal of the filter 110 is transmitted (step S101). That is, as for the two-dimensional signal, the region through which the two-dimensional signal of the filter 110 is transmitted is provided with a portion that blocks at least a part of the two-dimensional signal as described above. That is, the two-dimensional signal is imaged in the pattern of the filter 110.

The detector 120 is imaged in the pattern of the filter 110 and detects the power of the two-dimensional signal that has passed through the filter (step S102).

Next, the reconstruction unit 130 determines whether or not the power observation in step S102, which is a predetermined number of observations, has been performed (step S103). When the number of observations has not reached a predetermined number (step S103: No), the position of the two-dimensional signal and the pattern of the filter 110 changes so that the positional relationship between the two-dimensional signal and the pattern of the filter 110 changes. The relationship is controlled to change (step S104). This control is performed, for example, by the pattern control unit 111. Further, in this case, the positional relationship between the two-dimensional signal and the pattern of the filter 110 is the same as any of the positional relationships between the pixels of the two-dimensional signal detector 15 and the pattern of the filter 110 when obtaining the transformation matrix D. Is controlled. Then, returning to step S102, the detector 120 detects the power of the two-dimensional signal.

When the number of observations has reached a predetermined number (step S103: Yes), the reconstruction unit 130 reconstructs the image based on the plurality of powers detected in step S102 (step S105). That is, the reconstruction unit 130 reconstructs the image using the transformation matrix D obtained in advance based on the power of the signal obtained through k observations by the detector 120.

As described above, the image pickup apparatus 100 in the present embodiment is detected by changing (when different) the positional relationship between the two-dimensional signal indicating the image and the distribution of the portion that blocks at least a part of the filter 110. Reconstruct the image based on the power of multiple 2D signals. The number of observations of power detected when the image is reconstructed by the image pickup apparatus 100 is smaller than the number of pixels of the reconstructed image. Further, the positional relationship between the two-dimensional signal and the distribution of the portion blocking at least a part of the filter 110 can be easily changed by, for example, moving the filter 110 as appropriate. Therefore, the image pickup apparatus 100 makes it possible to improve the speed of the image pickup process.

Further, the detector 120 included in the image pickup apparatus 100 may be a detector that does not detect the position information but detects only the power of the signal. Therefore, even when it is difficult to array the detectors for various reasons including price, such as a far-infrared detector, the image pickup device 100 can easily acquire a two-dimensional image for a signal in the wavelength band. Become.

(Modification example)
A modified example of the above-described imaging device 100 can be considered.

In the image pickup apparatus 100, a monochromatic image was reconstructed. However, the image pickup apparatus 100 may be a so-called multicolor apparatus, that is, an apparatus for reconstructing images for a plurality of wavelengths.

FIG. 9 shows an example of a configuration in which the image pickup apparatus reconstructs images for a plurality of wavelengths. In the example shown in FIG. 9, the image pickup apparatus 101 includes a detector 121 instead of the detector 120. The detector 121 is a sensor having a plurality of pixels. As the detector 121, for example, an array sensor in which pixels are arranged in a two-dimensional direction or a line sensor is used, but the type of the detector 121 is not particularly limited.

Each pixel of the detector 121 is configured to detect a different wavelength or wavelength band by providing a filter that selectively transmits a specific wavelength or the like.

In the image pickup apparatus 101, an element corresponding to the wavelength of the detector 121 is selected according to the wavelength of the image to be reconstructed. Then, the image is reconstructed based on the power of the plurality of two-dimensional signals detected by the selected element. By changing the selected element and repeatedly reconstructing the image, it is possible to reconstruct the image for a large number of wavelengths.

The transformation matrix D used when reconstructing the image is different for each wavelength. Therefore, the transformation matrix D is obtained in advance for each wavelength. When reconstructing the image, the transformation matrix D according to the wavelength is used. Then, by repeatedly reconstructing the images for different wavelengths, the images for a large number of wavelengths are reconstructed.

Further, the apparatus for reconstructing images for a plurality of wavelengths is not limited to the imaging apparatus 101. For example, when reconstructing an image of an object illuminated by a light source, the light source is provided with a spectroscopic mechanism, and the image pickup apparatus 100 reconstructs an image for each of a plurality of wavelengths, so that an image for a plurality of wavelengths is reconstructed. Can be reconstructed.

An image pickup device 101 or the like enables image reconstruction for a large number of wavelengths. That is, hyperspectral imaging becomes possible with the imaging device 101 or the like.

(Simulation example)
It was confirmed by simulation that the image was reconstructed by the image pickup apparatus 100 described above.

The image to be reconstructed was an image of 128 pixels both vertically and horizontally. The pattern of the filter 110 and the control of the position of the pattern were as shown in FIG. 7C. The signal was sampled every time the pattern was rotated 0.2 degrees.

Under such conditions, the original image and the image reconstructed when 10% (percentage) of the number of pixels are sampled from the original image are shown in FIGS. 10A and 10B, respectively. Comparing the original image with the reconstructed image, as shown in FIGS. 10A and 10B, the outline of the person, the brightness of the image, and some of the hat accessories are restored to the reconstructed image. Was confirmed.

As described above, assuming that sampling is performed every time the pattern of the filter 110 is rotated by 0.2 degrees, sampling is possible 1800 times for each rotation of the pattern. Further, assuming that sampling of 10% (percentage) of the number of pixels is required in the reconstruction of an image of 128 × 128 pixels, as described above, one image is displayed for each rotation of the pattern. The sampling required for reconstruction will be performed.

Therefore, assuming that the rotation speed of the pattern is 1800 rpm (number of times / minute), it is possible to reconstruct about 30 images per second. That is, it was confirmed that the image pickup apparatus 100 can improve the speed of the image pickup process accompanied by the reconstruction of the image.

The embodiments of the present disclosure are not limited to the above embodiments. The configuration of the embodiments of the present disclosure can be modified in various ways as can be understood by those skilled in the art. Also, the configurations in the embodiments can be combined with each other without departing from the scope of the present disclosure.

This application claims priority on the basis of Japanese application Japanese Patent Application No. 2018-052222 filed on March 20, 2018 and incorporates all of its disclosures herein.

100 Imaging device 110 Filter 111 Pattern control unit 112 Rotating axis 120 Detector 130 Reconstruction unit 11 Lens 12 Iris 13 Scattering medium 14 Signal generator 15 Two-dimensional signal detector

Claims (10)

A filter having a region through which a two-dimensional signal showing an image is transmitted, and a portion provided in the region for blocking at least a part of the two-dimensional signal.
A detector that detects the power of the two-dimensional signal that has passed through the filter, and
A reconstruction means for reconstructing an image represented by the two-dimensional signal based on a plurality of the powers detected when the positional relationship between the two-dimensional signal imaged on the filter and the distribution of the portion is different.
An imaging device comprising. A pattern control means for controlling the position of the filter so that the positional relationship changes is provided.
The imaging device according to claim 1. The pattern control means controls the position of the filter so that the positional relationship changes by rotating the filter at a position where the two-dimensional signal is imaged.
The imaging device according to claim 2. The pattern control means moves the filter in a direction intersecting the direction in which the two-dimensional signal is transmitted at a position where the two-dimensional signal is imaged, so that the position of the filter is changed so that the positional relationship is changed. Control,
The imaging device according to claim 2 or 3. The filter includes a non-transmissive portion that blocks the two-dimensional signal as a portion that blocks at least a part of the two-dimensional signal.
The imaging device according to any one of claims 1 to 4. The filter includes a semi-transmissive portion that transmits a part of the two-dimensional signal and blocks a part of the two-dimensional signal as a portion that blocks at least a part of the two-dimensional signal.
The imaging device according to any one of claims 1 to 5. The reconstructing means reconstructs the image by an ADMM (alternation direction method of multipliers) algorithm.
The imaging device according to any one of claims 1 to 6. The reconstructing means reconstructs the image by the ADMM algorithm using the transformation matrix obtained for each of the distribution of the site and the positional relationship.
The imaging device according to claim 7. The detector has a plurality of pixels having different wavelengths of the two-dimensional signal to be detected.
The reconstruction means reconstructs the image at each of the plurality of wavelengths based on the plurality of the powers detected for each of the plurality of wavelengths of the two-dimensional signal.
The imaging device according to any one of claims 1 to 8. The power of the two-dimensional signal transmitted through a filter having a region through which the two-dimensional signal showing an image is transmitted and having a portion for blocking at least a part of the two-dimensional signal in the region is detected.
The image represented by the two-dimensional signal is reconstructed based on a plurality of the powers detected when the positional relationship between the two-dimensional signal imaged on the filter and the distribution of the portion is different.
Imaging method.

Images